Self-Driving Car Engineer Nanodegree

Deep Learning

Project: Build a Traffic Sign Recognition Classifier

In this notebook, a template is provided for you to implement your functionality in stages, which is required to successfully complete this project. If additional code is required that cannot be included in the notebook, be sure that the Python code is successfully imported and included in your submission if necessary.

Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the iPython Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.

In addition to implementing code, there is a writeup to complete. The writeup should be completed in a separate file, which can be either a markdown file or a pdf document. There is a write up template that can be used to guide the writing process. Completing the code template and writeup template will cover all of the rubric points for this project.

The rubric contains "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. The stand out suggestions are optional. If you decide to pursue the "stand out suggestions", you can include the code in this Ipython notebook and also discuss the results in the writeup file.

Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. In addition, Markdown cells can be edited by typically double-clicking the cell to enter edit mode.

---

Step 0: Load The Data

# Load pickled data
import pickle

# TODO: Fill this in based on where you saved the training and testing data

training_file   = "data/train.p"
validation_file = "data/valid.p"
testing_file    = "data/test.p"

with open(training_file, mode='rb') as f:
    train = pickle.load(f)
with open(validation_file, mode='rb') as f:
    valid = pickle.load(f)
with open(testing_file, mode='rb') as f:
    test = pickle.load(f)
    
X_train, y_train = train['features'], train['labels']
X_valid, y_valid = valid['features'], valid['labels']
X_test, y_test = test['features'], test['labels']

---

Step 1: Dataset Summary & Exploration

The pickled data is a dictionary with 4 key/value pairs:

• 'features' is a 4D array containing raw pixel data of the traffic sign images, (num examples, width, height, channels).
• 'labels' is a 1D array containing the label/class id of the traffic sign. The file signnames.csv contains id -> name mappings for each id.
• 'sizes' is a list containing tuples, (width, height) representing the original width and height the image.
• 'coords' is a list containing tuples, (x1, y1, x2, y2) representing coordinates of a bounding box around the sign in the image. THESE COORDINATES ASSUME THE ORIGINAL IMAGE. THE PICKLED DATA CONTAINS RESIZED VERSIONS (32 by 32) OF THESE IMAGES

Complete the basic data summary below. Use python, numpy and/or pandas methods to calculate the data summary rather than hard coding the results. For example, the pandas shape method might be useful for calculating some of the summary results.

Provide a Basic Summary of the Data Set Using Python, Numpy and/or Pandas

### Replace each question mark with the appropriate value. 
### Use python, pandas or numpy methods rather than hard coding the results

# TODO: Number of training examples
n_train = X_train.shape[0]

# TODO: Number of validation examples
n_validation = X_valid.shape[0]

# TODO: Number of testing examples.
n_test =X_test.shape[0]

# TODO: What's the shape of an traffic sign image?
image_shape = X_train[0].shape

# TODO: How many unique classes/labels there are in the dataset.
n_classes = len(set(y_train))  #https://stackoverflow.com/a/12282286

print("Number of training examples =", n_train)
print("Number of testing examples =", n_test)
print("Image data shape =", image_shape)
print("Number of classes =", n_classes)

Number of training examples = 34799
Number of testing examples = 12630
Image data shape = (32, 32, 3)
Number of classes = 43

Include an exploratory visualization of the dataset

Visualize the German Traffic Signs Dataset using the pickled file(s). This is open ended, suggestions include: plotting traffic sign images, plotting the count of each sign, etc.

The Matplotlib examples and gallery pages are a great resource for doing visualizations in Python.

NOTE: It's recommended you start with something simple first. If you wish to do more, come back to it after you've completed the rest of the sections. It can be interesting to look at the distribution of classes in the training, validation and test set. Is the distribution the same? Are there more examples of some classes than others?

### Data exploration visualization code goes here.
### Feel free to use as many code cells as needed.
import matplotlib.pyplot as plt
# Visualizations will be shown in the notebook.
%matplotlib inline

import matplotlib.lines as mlines
import matplotlib.markers as mmarkers
import random
import pandas as pd
import numpy as np
import textwrap
from PIL     import Image
from os      import listdir
from os.path import isfile, join, basename

Helper Classes and Functions

Classes to help in exploring and viewing data:

random.seed()

class Logger:
    '''
    Simple logger for project to make output messages look clean"
    '''
    
    def __init__(self, name, owner=None):
        if owner is not None:
            # https://stackoverflow.com/a/7983848
            name = name + "-" + str(id(owner))[-4:]
        
        self.name = name
        return
        
    def log(self, msg):
        print (self.name + " | " + msg)
        return


class SignNames_CSV:
    '''
    Provides access to the signnames.csv to retrieve sign names.
    '''

    def __init__(self):
        self.df = pd.read_csv("signnames.csv")
        return
    
    def signName(self, class_id):
        return self.df.at[class_id, 'SignName']
    
signnames_csv = SignNames_CSV()


def signName(class_id):
    '''
    Returns name of sign given class_id.
    
    Params:
    - class_id: integer from 0-42 representing id number of sign name
    
    Notes: 
    - Example: signName(0) retturns "Speed limit (20km/h)".
    - Use to map dataset label to sign name
    '''
    return signnames_csv.signName(class_id)


class TrafficSignsDataset:
    '''
    Wrapper for the training, validation, and testing datasets.
    '''
    
    def __init__(self, *args):
        
        self.logger = Logger("TrafficSignsDataset", self)
        
        if len(args) == 3:
            # initialized with imags, labels and n_classes
            images, labels, n_classes = args
        else:
            # len args == 2
            # initializes with folder_name and n_classes
            folder_name, n_classes = args
            images, labels = self.loadFolder(folder_name)
            
        self.org_X = images
        self.org_y = labels
            
        self.X = self.org_X.copy()
        self.y = self.org_y.copy()
        
        self.n_classes = n_classes
        
        return

    def loadFolder(self, folder_name):
    
        msg = "Loading signs from {}..."
        self.logger.log(msg.format(folder_name))

        pathnames = [join(folder_name, f) 
                     for f in listdir(folder_name) if isfile(join(folder_name, f))]

        n = 0
        images = []
        labels = []

        for pathname in pathnames:
            
            filename = basename(pathname) 
            classId = filename.split(",")[0]

            pil_img = Image.open(pathname).convert('RGB')
            
            pil_img20x20 = pil_img.resize((20, 20), Image.BILINEAR)
            img = np.pad (np.array(pil_img20x20), [(6, 6), (6, 6), (0, 0)])
            
            '''
            #if not padding, use:
            pil_img32x32 = pil_img.resize((32, 32), Image.BILINEAR)
            img = np.array(pil_img32x32)
            '''

            labels.append(classId)
            images.append(img)

            msg = "Loaded {} with label {}."
            self.logger.log(msg.format(filename, classId))

            n += 1

        msg = "...{} signs loaded."
        self.logger.log (msg.format(n))
        
        return images, labels
    
    def data(self):
        return (self.X, self.y, self.n_classes)
    
    def processImages(self, img_processors):
        
        self.logger.log("Processing images...")
        
        for img_processor in img_processors:
            
            msg = "Applying {}() to images."
            self.logger.log(msg.format(img_processor.__name__))
            
            self.X = [img_processor(img) for img in self.X]
            
        self.logger.log("...images processed.")            
        
        return
    
    def addFakes (self, img_processors):
        
        self.logger.log("Adding fake images...")
        max_samples = np.max(np.bincount(self.y)) * 3
        
        total_added = 0
        
        for classId in range(self.n_classes):

            x = self.X[self.y == classId]
            y = self.y[self.y == classId]

            n_x = x.shape[0]

            aug_x = np.array([])
            aug_y = np.array([])

            if n_x < max_samples:

                n_repeats = (max_samples - n_x) // n_x
                n_remain = (max_samples - n_x) % n_x

                aug_x = np.repeat(x, n_repeats, axis=0)
                aug_x = np.concatenate ((aug_x, x[:n_remain]))
                for img_processor in img_processors:
                    aug_x = np.array([img_processor(img) for img in aug_x])

                aug_y = np.repeat(y, n_repeats, axis=0)
                aug_y = np.concatenate ((aug_y, y[:n_remain]))
                
                samples_added = max_samples - n_x
                msg = "Added {} fake images for classId {}."
                self.logger.log(msg.format(samples_added, classId))
                
                total_added += samples_added
                
            if aug_x.shape[0] > 0:
                self.X = np.concatenate((self.X, aug_x))
                self.y = np.concatenate((self.y, aug_y))
                
        self.X = np.array(self.X, dtype=np.uint8)
        
        msg = "...{:,} fake images added."
        self.logger.log (msg.format(total_added))
        
        msg= "{:,} samples available for training."
        self.logger.log (msg.format(self.X.shape[0]))

        return
    
    def restoreOrgImages(self):
        
        self.X = np.array(self.org_X, dtype=np.uint8)
        self.y = np.array(self.org_y)
        
        self.logger.log("Original images and labels restored.")
        
        return
    
class Viewer:
    '''
    Provides methods for plotting data on notebook.
    '''
    
    def __init__(self):
        return
    
    def imShowTrainingSigns(self, trainingSigns):

        ds = list(zip(trainingSigns.y, trainingSigns.X))
        df = pd.DataFrame(data=ds, columns=['Label', 'Image'])

        # https://stackoverflow.com/a/27430940
        nrows = trainingSigns.n_classes
        fig = plt.figure(figsize=(11, 2.0*nrows))

        for classId in range(trainingSigns.n_classes):
            ax = fig.add_subplot(trainingSigns.n_classes, 1, classId + 1)
            ax.set_title(signName(classId))
            ax.axis('off')
            
        n_samples = 8
        for classId in range(trainingSigns.n_classes):
            df_sel = df.loc[(df['Label'] == classId), 'Image'].sample(n_samples)
            for i in range(n_samples):
                ax = fig.add_subplot(trainingSigns.n_classes, n_samples, classId*n_samples + i + 1)
                ax.axis('off')
                ax.imshow(df_sel.iloc[i])
        plt.show()    
        
        return
    
    def imShowTrainingSigns5x5(self, trainingSigns):
        fig = plt.figure(figsize=(11,11))
        num_rows = 5
        num_cols = 5
        for i in range(num_cols*num_rows):
            img = random.choice(trainingSigns.X)
            ax = plt.subplot(num_rows, num_cols, i + 1)
            ax.imshow(img)
        plt.show()
        return
    
    def imShowImages(self, images, titles=None):
        
        n_images = len(images)
        n_cols = 5
        n_rows = n_images//n_cols + 1
        plt.figure(figsize=(11, (2.5 * n_rows)))
        for i, img in enumerate(images):
            ax = plt.subplot(n_rows, n_cols, i + 1)
            ax.imshow(img)
            if titles is not None:
                ax.set_title(titles[i])
            
        return

    def barhPctBySigns(self, trainingSigns, validSigns, testSigns):
        y_train = trainingSigns.y
        y_valid = validSigns.y
        y_test  = testSigns.y
        
        bincount_y_train = np.bincount(y_train)
        bincount_y_valid = np.bincount(y_valid)
        bincount_y_test  = np.bincount(y_test)
        # bin_count[i] = num samples for sign i (where i=classId)

        pct_y_train = bincount_y_train / sum(bincount_y_train)
        pct_y_valid = bincount_y_valid / sum(bincount_y_valid)
        pct_y_test  = bincount_y_test  / sum(bincount_y_test)
        # pct_y_train[i] = percentage of total samples of sign i 

        pct_y_train_sorted_idx = np.argsort(pct_y_train)
        # pct_y_train_sorted_idx an array of sign classIds sorted by their percentage

        # ref: https://stackoverflow.com/a/59421062
        bar_wid = 0.3

        y_pos      = np.arange(len(pct_y_train))
        pcts       = [pct_y_train[class_id] for class_id in pct_y_train_sorted_idx]
        pcts_valid = [pct_y_valid[class_id] for class_id in pct_y_train_sorted_idx]
        pcts_test  = [pct_y_test[class_id] for class_id in pct_y_train_sorted_idx]
        y_labels   = [textwrap.fill(signName(class_id), width=25) 
                      for class_id in pct_y_train_sorted_idx]

        fig, ax = plt.subplots(figsize=(11, 35))

        ax.barh(y_pos + 2*bar_wid, pcts,       bar_wid, label="Training")
        ax.barh(y_pos + bar_wid,   pcts_valid, bar_wid, label="Valid")
        ax.barh(y_pos,             pcts_test,  bar_wid, label="Test")

        ax.set_yticks(y_pos  + bar_wid)
        ax.set_yticklabels(y_labels)
        ax.set_xlabel('Percentage of Sample')
        ax.set_title("Percentage of Sample by Sign and Dataset")

        plt.legend(loc='best')
        plt.margins(y=0)
        plt.show()

        return
    
    def barhCountTrainingSigns(self, trainingSigns):
        sign_count_y_train = np.bincount(trainingSigns.y)

        # ref: https://stackoverflow.com/a/59421062
        y_pos        = np.arange(len(sign_count_y_train))
        counts_train = [sign_count_y_train[class_id] 
                        for class_id in range(trainingSigns.n_classes)]
        
        y_labels     = [textwrap.fill(signName(class_id), width=25) 
                        for class_id in range(trainingSigns.n_classes)]

        fig, ax = plt.subplots(figsize=(11, 35))

        # ax.barh(y_pos, counts_train)
        ax.barh(y_pos, counts_train)
        ax.set_yticks(y_pos)
        ax.set_yticklabels(y_labels)
        ax.invert_yaxis()
        ax.set_xlabel('Number of Training Samples')
        ax.set_title("Number of Training Samples by Sign")

        plt.margins(y=0)
        plt.show()

        return        
    
    def imShowPredictions(self, model_name, images, predictions):
        
        num_images = len(images)
        
        fig, axs = plt.subplots(num_images, 2, figsize=(15, 3.5*num_images))
        if num_images == 1:
            axs = [axs]
        
        for ax, img, vals, idxs in zip(axs, images, predictions.values, predictions.indices):
            
            pcts  = [100*v for v in vals]
            signs = [textwrap.fill(signName(i), width=25) for i in idxs]
            
            # https://matplotlib.org/gallery/lines_bars_and_markers/barh.html#sphx-glr-gallery-lines-bars-and-markers-barh-py
            img_ax = ax[0]
            bar_ax = ax[1]

            img_ax.imshow(img)
            img_ax.set_title("Image of Sign")
            
            ypos = np.arange(len(signs))
            bar_ax.barh(ypos, pcts)
            bar_ax.set_yticks(ypos)
            bar_ax.set_yticklabels(signs)
            bar_ax.invert_yaxis()
            bar_ax.set_xlabel("Confidence")
            bar_ax.set_title("Classification (Top 5)")
            bar_ax.set_xlim([0,100])

        fig.suptitle(model_name + " Classification of some German Traffic Signs", y = 1.0)
        fig.tight_layout()
        plt.show()    
        
        return
    
    def imShowConvOutputs(self, conv_outputs):

        # layer_ouputs in shape (1, ?, ?, depth)
        features = layer_outputs[0]
        depth = layer_outputs.shape[3]
        
        plt.figure(figsize=(15, (2.5 * (depth//8 + 1))))

        n_cols = 8
        n_rows = depth//n_cols + 1
        
        for level in range(depth):
            
            img = features[:,:,level] * 255
            
            plt.subplot(n_rows, n_cols, level + 1)
            plt.imshow(img, 
                       interpolation='nearest', 
                       cmap='gray',
                       vmin=np.amin(img), vmax=np.amax(img))
            
        return
    
    def plotAccHist(self, model_name, acc_hist):
        
        epochs = [i + 1 for i in range(len(acc_hist))]
        
        idx_max_acc = np.argmax(acc_hist)
        max_acc = acc_hist[idx_max_acc] * 100
        title = model_name + " max validation accuracy {:.1f}% at epoch {}."
        
        # https://stackoverflow.com/a/47392973
        max_acc_caret = mlines.Line2D([], [], color='r', 
                              marker='D', 
                              linestyle='None',
                              label='Max Validation Accuracy')
        
        fig = plt.figure(figsize=(10, 5))
                     
        ax = fig.add_subplot(1, 1, 1)
        ax.set_title(title.format(max_acc, idx_max_acc + 1))
        ax.set_xlabel("Epoch")
        # ax.set_xticks(epochs)
        ax.set_ylabel("Validation Accuracy")
        ax.plot(epochs, acc_hist, 
                marker='D', mec='r', mfc='r', markevery=[idx_max_acc], markersize=8)
        ax.legend(handles=[max_acc_caret], loc='lower right')
        ax.set_xlim([1, None])
        ax.set_ylim(top=1)
                     
        plt.show()
        
        return
    
    def barhScores(self, title, scores, y_labels):
        
        # ref: https://stackoverflow.com/a/59421062
        y_pos    = np.arange(len(scores))

        fig, ax = plt.subplots(figsize=(11, 3))

        ax.barh(y_pos, scores)
        ax.set_yticks(y_pos)
        ax.set_yticklabels(y_labels)
        ax.set_xlabel('Score')
        ax.set_title(title)
        
        for i, score in enumerate(scores):
            ax.text(score, i, "{:.3f}".format(score))

        ax.invert_yaxis()
            
        plt.margins(y=0)
        plt.show()
        
        return
    
    def barhClassScores(self, title, scores, n_lowest=None):
        
        if n_lowest is None:
            n_lowest = len(scores)
            
        sorted_idx = np.argsort(scores)[:n_lowest]

        sorted_scores = [scores[i] for i in sorted_idx]
        y_labels      = [textwrap.fill(signName(i), width=25) for i in sorted_idx]
        y_pos         = np.arange(len(sorted_scores))

        fig, ax = plt.subplots(figsize=(11, 0.5*len(sorted_scores)))
        

        ax.barh(y_pos, sorted_scores)
        ax.set_yticks(y_pos)
        ax.set_yticklabels(y_labels)
        ax.set_xlabel('Score')
        ax.set_title(title)
        
        for i, score in enumerate(sorted_scores):
            ax.text(score, i, "{:.3f}".format(score))

        ax.invert_yaxis()

        plt.margins(y=0)
        plt.show()
        
        return
    
    def barhCombinedClassScores(self, title, accuracy, precision, recall, rank=-10):
        
        avg_scores = np.average([accuracy, precision, recall], axis=0)
        if rank < 0:
            sorted_idx = np.argsort(avg_scores)[:-rank]
        else:
            sorted_idx = np.argsort(avg_scores)[-rank:]            
        
        sorted_acc    = [accuracy[i]  for i in sorted_idx]
        sorted_pre    = [precision[i] for i in sorted_idx]
        sorted_rcl    = [recall[i]    for i in sorted_idx]
        y_labels      = [textwrap.fill(signName(i), width=25) for i in sorted_idx]
        y_pos         = np.arange(len(y_labels))
        
        bar_wid = 0.3
        fig, ax = plt.subplots(figsize=(11, 1.5*len(y_labels)))
        
        ax.set_xlim(0, 1.3)
        
        ax.barh(y_pos,             sorted_acc,  bar_wid, label="Accuracy")
        ax.barh(y_pos + bar_wid,   sorted_pre,  bar_wid, label="Precision")
        ax.barh(y_pos + 2*bar_wid, sorted_rcl,    bar_wid, label="Recall")
        
        for i, score in enumerate(sorted_acc):
            ax.text(score, i, "{:.3f}".format(score))
        for i, score in enumerate(sorted_pre):
            ax.text(score, i + bar_wid, "{:.3f}".format(score))
        for i, score in enumerate(sorted_rcl):
            ax.text(score, i + 2*bar_wid, "{:.3f}".format(score))
        
        ax.set_yticks(y_pos + bar_wid)
        ax.set_yticklabels(y_labels)
        
        ax.set_title(title)
        
        if rank < 0:
            ax.invert_yaxis()
        ax.legend(loc='upper right')

        plt.xticks([])
        plt.margins(y=0)
        plt.show()
        
        return
        
        

Instantiate Datasets and Viewer

trainingSigns = TrafficSignsDataset(X_train, y_train, n_classes)
validSigns    = TrafficSignsDataset(X_valid, y_valid, n_classes)
testSigns     = TrafficSignsDataset(X_test, y_test, n_classes)
vwr           = Viewer()

Explore Some Traffic Signs

View signs in trainingSigns dataset sorted by type.

• Some are really dark and will need brightness adjusted (in later cells).
• Double curve sign is a double curve to the left (No double curve to the right).
• A quick look on wikipedia shows there are actually more than the 43 types of signs in this dataset.
• Color, shape, and orientation matter.

vwr.imShowTrainingSigns(trainingSigns)

View Samples Sizes by Training Sign

• Proportions by dataset (training, validation, test) are about the same across all signs.
• Over half the signs appear to have signficantly less than other signs. Example: "Beware of Ice and snow" is much less than "Speed limit (50km/h)"
• Additional training signs will be generated so model training has a better chance to be exposed to all types of signs.

vwr.barhPctBySigns(trainingSigns, validSigns, testSigns)

---

Step 2: Design and Test a Model Architecture

Design and implement a deep learning model that learns to recognize traffic signs. Train and test your model on the German Traffic Sign Dataset.

The LeNet-5 implementation shown in the classroom at the end of the CNN lesson is a solid starting point. You'll have to change the number of classes and possibly the preprocessing, but aside from that it's plug and play!

With the LeNet-5 solution from the lecture, you should expect a validation set accuracy of about 0.89. To meet specifications, the validation set accuracy will need to be at least 0.93. It is possible to get an even higher accuracy, but 0.93 is the minimum for a successful project submission.

There are various aspects to consider when thinking about this problem:

• Neural network architecture (is the network over or underfitting?)
• Play around preprocessing techniques (normalization, rgb to grayscale, etc)
• Number of examples per label (some have more than others).
• Generate fake data.

Here is an example of a published baseline model on this problem. It's not required to be familiar with the approach used in the paper but, it's good practice to try to read papers like these.

Pre-process the Data Set (normalization, grayscale, etc.)

Minimally, the image data should be normalized so that the data has mean zero and equal variance. For image data, (pixel - 128)/ 128 is a quick way to approximately normalize the data and can be used in this project.

Other pre-processing steps are optional. You can try different techniques to see if it improves performance.

Use the code cell (or multiple code cells, if necessary) to implement the first step of your project.

### Preprocess the data here. It is required to normalize the data. Other preprocessing steps could include 
### converting to grayscale, etc.
### Feel free to use as many code cells as needed.

import numpy as np
import cv2
import skimage as sk
from skimage.transform import rotate

Helper Functions for Pre-processing and Generating Additional Training Samples (Fake Data)

• equalizeHist() will brighten dark images
• Global variable image_processors is a list of the pre-processor functions applied to datasets before use in training and testing model.
• Other functions such as shift(), zoom(), etc used in generating additional training samples.

def RGB_to_norm(img):
    return (np.float32(img) - 128)/128

def norm_to_RGB(norm_img):
    return (np.uint8(128 * norm_img) + 128)

def equalizeHist(orgimg, v_thresh=128):
    '''
    Brightens dark images.
    
    Params:
    - orgimg: original image (RGB)
    - v_thresh: max integer of the average value of the image for brightening to occur
    '''
    hsv = cv2.cvtColor(orgimg, cv2.COLOR_RGB2HSV)
    mean_v = np.mean(hsv[:,:,2])
    if mean_v < v_thresh:
        equ = cv2.equalizeHist(hsv[:,:,2])
        hsv[:,:,2] = equ
        img = cv2.cvtColor(hsv, cv2.COLOR_HSV2RGB)
    else:
        img = orgimg
    return img

image_processors = [equalizeHist, RGB_to_norm]

# https://medium.com/@thimblot/data-augmentation-boost-your-image-dataset-with-few-lines-of-python-155c2dc1baec

def shift(image, d=None):
    '''
    Shifts image left, right, up or down from 1 to 6 pixels. 
    Used in generating additional training samples.
    '''
    if d is None:
        d = random.randint(1, 6)
    direction = random.choice([1, 2, 3, 4])
    if direction == 1:
        image[:-d] = image[d:]
    elif direction == 2:
        image[d:] = image[:-d]
    elif direction == 3:
        image[:,d:] = image[:,:-d]
    else:
        image[:,:-d] = image[:,d:]
        
    return image

def crop(image, size=32):
    '''
    Crops image to 32x32. Used after image is zoomed in.
    '''
    sizes = np.array(image.shape[:2]) - 32
    lower =  sizes // 2
    upper = image.shape[:2] - (lower + (sizes % 2))
    img = image[lower[0]:upper[0], lower[1]:upper[1]]
    return img

def zoom(image, scale=None):
    '''
    Zooms in on an image from 1.1x to 1.6x. Uses crop to ensure img is 32x32
    Used in generating additional training samples.
    '''
    if scale is None:
        scale = random.uniform(1.1, 1.6)
    img = sk.transform.rescale(image, scale, multichannel=True, preserve_range=True).astype(np.uint8)
    return crop(img)

def rotate (image, deg=None):
    '''
    Rotates image from -15 to 15 degrees.
    Used in generating additional training samples.
    '''
    if deg is None:
        deg = random.uniform(-15, 15)
    return sk.transform.rotate(image, deg, preserve_range=True).astype(np.uint8)

def preserve(image):
    '''
    Use same image.
    Used in generating additional training samples.
    '''
    return image.astype(np.uint8)

def fakeify(image):
    '''
    Generates fake images to increase number of training samples.
    
    Notes:
    - Randomly selects from preserve, rotate, zoom, shift to generate image.
    '''
    f = random.choices ([preserve, rotate, zoom, shift], weights=[1, 3, 3, 3])[0]
    return f(image)

Examples of Brightening and Fake Generation

trainingSigns.restoreOrgImages()

org_image = trainingSigns.X[1000].copy()
brt_image = equalizeHist(org_image.copy())
rot_image = rotate(brt_image.copy(), 15)
zm_image  = zoom(brt_image.copy(), 1.6)
shf_image = shift(brt_image.copy(), 6)

images = [org_image, brt_image, rot_image, zm_image, shf_image]
titles  = ["Original", "Brightened", "Rotated", "Zoomed", "Shifted"]

vwr.imShowImages(images, titles)

TrafficSignsDataset-2280 | Original images and labels restored.

Generate Additional Training Samples

A total of 259,290 samples are used for training after adding additional training samples.

trainingSigns.restoreOrgImages()
trainingSigns.addFakes([fakeify])

TrafficSignsDataset-2280 | Original images and labels restored.
TrafficSignsDataset-2280 | Adding fake images...
TrafficSignsDataset-2280 | Added 5850 fake images for classId 0.
TrafficSignsDataset-2280 | Added 4050 fake images for classId 1.
TrafficSignsDataset-2280 | Added 4020 fake images for classId 2.
TrafficSignsDataset-2280 | Added 4770 fake images for classId 3.
TrafficSignsDataset-2280 | Added 4260 fake images for classId 4.
TrafficSignsDataset-2280 | Added 4380 fake images for classId 5.
TrafficSignsDataset-2280 | Added 5670 fake images for classId 6.
TrafficSignsDataset-2280 | Added 4740 fake images for classId 7.
TrafficSignsDataset-2280 | Added 4770 fake images for classId 8.
TrafficSignsDataset-2280 | Added 4710 fake images for classId 9.
TrafficSignsDataset-2280 | Added 4230 fake images for classId 10.
TrafficSignsDataset-2280 | Added 4860 fake images for classId 11.
TrafficSignsDataset-2280 | Added 4140 fake images for classId 12.
TrafficSignsDataset-2280 | Added 4110 fake images for classId 13.
TrafficSignsDataset-2280 | Added 5340 fake images for classId 14.
TrafficSignsDataset-2280 | Added 5490 fake images for classId 15.
TrafficSignsDataset-2280 | Added 5670 fake images for classId 16.
TrafficSignsDataset-2280 | Added 5040 fake images for classId 17.
TrafficSignsDataset-2280 | Added 4950 fake images for classId 18.
TrafficSignsDataset-2280 | Added 5850 fake images for classId 19.
TrafficSignsDataset-2280 | Added 5730 fake images for classId 20.
TrafficSignsDataset-2280 | Added 5760 fake images for classId 21.
TrafficSignsDataset-2280 | Added 5700 fake images for classId 22.
TrafficSignsDataset-2280 | Added 5580 fake images for classId 23.
TrafficSignsDataset-2280 | Added 5790 fake images for classId 24.
TrafficSignsDataset-2280 | Added 4680 fake images for classId 25.
TrafficSignsDataset-2280 | Added 5490 fake images for classId 26.
TrafficSignsDataset-2280 | Added 5820 fake images for classId 27.
TrafficSignsDataset-2280 | Added 5550 fake images for classId 28.
TrafficSignsDataset-2280 | Added 5790 fake images for classId 29.
TrafficSignsDataset-2280 | Added 5640 fake images for classId 30.
TrafficSignsDataset-2280 | Added 5340 fake images for classId 31.
TrafficSignsDataset-2280 | Added 5820 fake images for classId 32.
TrafficSignsDataset-2280 | Added 5431 fake images for classId 33.
TrafficSignsDataset-2280 | Added 5670 fake images for classId 34.
TrafficSignsDataset-2280 | Added 4950 fake images for classId 35.
TrafficSignsDataset-2280 | Added 5700 fake images for classId 36.
TrafficSignsDataset-2280 | Added 5850 fake images for classId 37.
TrafficSignsDataset-2280 | Added 4170 fake images for classId 38.
TrafficSignsDataset-2280 | Added 5760 fake images for classId 39.
TrafficSignsDataset-2280 | Added 5730 fake images for classId 40.
TrafficSignsDataset-2280 | Added 5820 fake images for classId 41.
TrafficSignsDataset-2280 | Added 5820 fake images for classId 42.
TrafficSignsDataset-2280 | ...224,491 fake images added.
TrafficSignsDataset-2280 | 259,290 samples available for training.

View Some of the Traffic Signs in Training Dataset

vwr.imShowTrainingSigns5x5(trainingSigns)

View Sample Sizes

After generating additional samples for training, the number of samples for each sign should be equal.

vwr.barhCountTrainingSigns(trainingSigns)

Pre-process Datasets

Brighten and normalize datasets.

trainingSigns.processImages(image_processors)

validSigns.restoreOrgImages()
validSigns.processImages(image_processors)

testSigns.restoreOrgImages()
testSigns.processImages(image_processors)

TrafficSignsDataset-2280 | Processing images...
TrafficSignsDataset-2280 | Applying equalizeHist() to images.
TrafficSignsDataset-2280 | Applying RGB_to_norm() to images.
TrafficSignsDataset-2280 | ...images processed.
TrafficSignsDataset-2168 | Original images and labels restored.
TrafficSignsDataset-2168 | Processing images...
TrafficSignsDataset-2168 | Applying equalizeHist() to images.
TrafficSignsDataset-2168 | Applying RGB_to_norm() to images.
TrafficSignsDataset-2168 | ...images processed.
TrafficSignsDataset-0304 | Original images and labels restored.
TrafficSignsDataset-0304 | Processing images...
TrafficSignsDataset-0304 | Applying equalizeHist() to images.
TrafficSignsDataset-0304 | Applying RGB_to_norm() to images.
TrafficSignsDataset-0304 | ...images processed.

Model Architecture

### Define your architecture here.
### Feel free to use as many code cells as needed.

import tensorflow as tf

from tensorflow.contrib.layers import flatten
from sklearn.utils             import shuffle
from PIL                       import Image
from datetime                  import datetime

WARNING: Logging before flag parsing goes to stderr.
W0211 22:24:59.330260  5528 __init__.py:329] Limited tf.compat.v2.summary API due to missing TensorBoard installation.

Helper Functions to Define Model Layers

Functions below help to create convolutional, pooling and fully connected layers.

def conv2D(x, output_shape, s=1, mu=0, sigma=0.1, activation=tf.nn.relu):
    
    input_shape = x.shape.as_list()[1:]
    
    input_depth  = input_shape[2]
    output_depth = output_shape[2]
    
    stride = [1, s, s, 1]
    
    # size of filter (length=Width)
    # ref: lesson 12: Solution: Convolution Output Shape for "VALID" padding
    f_ht  = input_shape[0] - output_shape[0]*s + 1
    f_wid = input_shape[1] - output_shape[1]*s + 1
    
    # shape of weights
    w_shape = (f_ht, f_wid, input_depth, output_depth)
    
    w = tf.Variable(tf.truncated_normal(w_shape, mu, sigma))
    b = tf.Variable(tf.zeros(output_depth))
    c = tf.nn.conv2d(x, w, stride, padding='VALID') + b
    a = activation(c)
    
    return a


def pooling(x, output_shape, s=2, pool=tf.nn.max_pool):
    
    input_shape = x.shape.as_list()[1:]
    input_depth = input_shape[2]
    
    # size of filter
    # ref: lesson 23: Quiz: Pooling Mechanics
    f_ht  = input_shape[0] - (output_shape[0] - 1) * s
    f_wid = input_shape[1] - (output_shape[1] - 1) * s
    
    fil    = [1, f_ht, f_wid, 1]
    stride = [1, s, s, 1]
    
    p = pool(x, fil, stride, padding='VALID')
    
    return p


def connected(x, output_shape,  mu=0, sigma=0.1, activation=tf.nn.relu):
    input_shape = x.shape.as_list()[1]
    output_shape = output_shape[0]
    w_shape = (input_shape, output_shape)
    
    w = tf.Variable(tf.truncated_normal(w_shape, mu, sigma))
    b = tf.Variable(tf.zeros(output_shape))
    z = tf.add(tf.matmul(x, w), b)
    
    if activation is None:
        a = z
    else:
        a = activation(z)
    
    return a

Class Wrappers for Layers

Classes below wrap the helper functions above. A Keras-inspired convention is used in creating layers and models.

class Layer():
    def __init__(self, output_shape=None):
        self.output_shape = output_shape
        self.tensor = None
        self.model = None
        self.setName()
        return
    
    def setName(self):
        self.name = "Layer"
        return
    
    def summary(self):
        return self.name.ljust(10) + ":" + str(self.tensor.shape)
    
    def connect(self, *prev_layers):
        self.tensor = prev_layers[0].tensor
        return self
    
    def owner(self, model = None):
        if model is not None:
            self.model = model
        return self.model

    
class Input(Layer):
    
    def __init__(self, X):
        
        super().__init__(self)
        self.tensor = X
        
    def setName(self):
        self.name = "Input"
        return
    
    
class Conv2D(Layer):

    def setName(self):
        self.name = "Conv2D"
        return
    
    def connect(self, *prev_layer):
        self.model = prev_layer[0].model
        self.tensor = conv2D(prev_layer[0].tensor, self.output_shape)
        return self
    

class Pooling(Layer):
    
    def setName(self):
        self.name = "Pooling"
        return
    
    def connect(self, *prev_layer):
        self.model = prev_layer[0].model
        self.tensor = pooling(prev_layer[0].tensor, self.output_shape)
        return self
    
    
class Dropout(Layer):
    
    def setName(self):
        self.name = "Dropout"
        return

    def connect(self, *prev_layer):
        self.model = prev_layer[0].model
        self.keep_prob = self.model.keep_prob
        self.tensor = tf.nn.dropout(prev_layer[0].tensor, self.keep_prob)
        return self
    

class Flatten(Layer):
    
    def setName(self):
        self.name = "Flatten"
        return

    def connect(self, *prev_layer):
        self.model = prev_layer[0].model
        self.tensor = flatten(prev_layer[0].tensor)
        return self
    
    
class Connected(Layer):

    def setName(self):
        self.name = "Connected"
        return
    
    def connect(self, *prev_layer):
        self.model = prev_layer[0].model
        self.tensor = connected(prev_layer[0].tensor, self.output_shape)
        return self

Class for Defining a Deep Neural Network

A base Model class and Sequential Model Keras-inspired classed are defined below. The base Model class is used to construct a multi-scale network then Optional section of this project. The more traditional Sequential is used to construct the main classifier for this project.

class Model:
    
    def __init__(self, name, input_shape, n_classes):
        self.logger = Logger('Model', self)
        
        self.name = name
        
        x_shape = (None, input_shape[0], input_shape[1], input_shape[2])
        self.x  = tf.placeholder(tf.float32, x_shape)
        
        self.input_layer = Input(self.x)
        self.input_layer.owner(self)
        
        self.y         = tf.placeholder(tf.int32, (None))
        self.keep_prob = tf.placeholder(tf.float32)

        self.n_classes    = n_classes
        self.logits       = None
        self.oh_labels    = None
        self.mean_loss    = None
        self.tot_correct  = None
        self.acc_history  = None

        self.saver = None 
        
        return
    
    def inputLayer(self):
        return self.input_layer
    
    def connectLogits(self, prev_layer):
    
        self.logits = connected(prev_layer.tensor, [self.n_classes], activation=None)

        self.oh_labels = tf.one_hot(self.y, self.n_classes)
        losses         = tf.nn.softmax_cross_entropy_with_logits(labels=self.oh_labels, 
                                                                 logits=self.logits)
        self.mean_loss = tf.reduce_mean(losses)
        
        correct_predictions = tf.equal(tf.argmax(self.oh_labels, 1), tf.argmax(self.logits, 1))
        self.tot_correct   = tf.reduce_sum(tf.cast(correct_predictions, tf.float32))
        
        return
    
    def accuracy(self, data, batch_size):
        '''
        calculates accuracy without using tf.metrics.accuracy
        '''
        
        X_data = data[0]
        y_data = data[1]
        
        num_examples = len(X_data)
        total_correct = 0
        sess = tf.get_default_session()        
        
        for offset in range(0, num_examples, batch_size):
            
            batch_x   = X_data[offset:offset+batch_size] 
            batch_y   = y_data[offset:offset+batch_size]
            feed_dict = {self.x: batch_x, 
                         self.y: batch_y, 
                         self.keep_prob: 1.0}
            
            batch_correct = sess.run(self.tot_correct, feed_dict=feed_dict)
            total_correct += batch_correct
            
        return round (total_correct / num_examples, 3)
    
    def measure(self, tf_metrics, data, classId=None, batch_size=128):
        
        sess = tf.get_default_session()        

        X_data = data[0]
        y_data = data[1]
        
        # ref!!! http://ronny.rest/blog/post_2017_09_11_tf_metrics/
        # setup inputs to metric function
        if classId is None:
            labels      = tf.argmax(self.oh_labels, 1)
            predictions = tf.argmax(self.logits, 1)
        else:
            # used if metrics are by classid
            tf_classId  = tf.constant(classId, tf.int64)
            labels      = tf.equal(tf.argmax(self.oh_labels, 1), tf_classId)
            predictions = tf.equal(tf.argmax(self.logits, 1), tf_classId)
        
        op_metrics     = []
        op_upd_metrics = []
        for i, tf_metric in enumerate(tf_metrics):
            
            # setup tensorflow metrics precision tensor
            name = 'metric_' + str(i)
            op_metric, op_upd_metric = tf_metric(labels, predictions, name=name)
            op_metrics.append(op_metric) 
            op_upd_metrics.append(op_upd_metric)

            # initialize variables used by tf.metrics.precision
            local_vars       = tf.GraphKeys.LOCAL_VARIABLES
            metric_vars      = tf.get_collection(local_vars, scope=name)
            metric_vars_init = tf.variables_initializer(var_list=metric_vars)
            sess.run(metric_vars_init)
        
        n_samples = len(X_data)
        
        for offset in range(0, n_samples, batch_size):
            
            batch_x   = X_data[offset:offset+batch_size] 
            batch_y   = y_data[offset:offset+batch_size]
            feed_dict = {self.x        : batch_x, 
                         self.y        : batch_y, 
                         self.keep_prob: 1.0}
            
            for op_upd_metric in op_upd_metrics:
                sess.run(op_upd_metric, feed_dict=feed_dict)
        
        scores = [sess.run(op_metric) for op_metric in op_metrics]
        
        return scores
    
    def saveModel(self):
        
        sess = tf.get_default_session()
        self.saver.save(sess, "checkpoints/" + self.name)
        msg = "{} saved."
        self.logger.log(msg.format(self.name))
        
        return
    
    def afterEpoch(self, epoch, elapsed, acc, acc_save, hi_acc):
        self.acc_history.append(acc)
        elapsed += 1
        if acc > acc_save:
            if acc > hi_acc:
                elapsed = 0    # restart count
                hi_acc = acc
                msg = "Epoch {:4d} - New High Acc: {:.3f}"
                self.logger.log(msg.format(epoch + 1, hi_acc))
                
                self.saveModel()
            else:
                msg = "Epoch {:4d} - Acc: {:.3f} - Highest: {:.3f}"
                self.logger.log(msg.format(epoch + 1, acc, hi_acc))
        elif hi_acc > 0:
            msg = "Epoch {:4d} -  Acc: {:.3f} - Highest: {:.3f}"
            self.logger.log(msg.format(epoch + 1, acc, hi_acc))
        else:
            msg = "Epoch {:4d} - Acc: {:.3f}"
            self.logger.log(msg.format(epoch + 1, acc))

        epoch += 1
            
        return epoch, elapsed, hi_acc
    
    def train(self, training_data, validation_data, epochs_done, batch_size, lr=0.001, 
              acc_save=0.93, acc_done=0.982, keep_prob=1.0, ):
        
        X_train = training_data[0]
        y_train = training_data[1]
        
        optimizer = tf.train.AdamOptimizer(learning_rate=lr)
        minimizer = optimizer.minimize(self.mean_loss)
        
        self.saver = tf.train.Saver()
        
        self.logger.log("Training...")
        
        start_time = datetime.now() # tip from https://stackoverflow.com/a/45668747
        self.logger.log("Timer started.")
        
        with tf.Session() as sess:
            try:
            
                sess.run(tf.global_variables_initializer())

                num_examples = len(X_train)

                self.acc_history = []
                hi_acc = 0
                epoch = 0
                elapsed = 0
                acc = 0
                while (elapsed < epochs_done) and (acc < acc_done):

                    X_train, y_train = shuffle(X_train, y_train)

                    for offset in range(0, num_examples, batch_size):
                        end       = offset + batch_size
                        batch_x   = X_train[offset:end] 
                        batch_y   = y_train[offset:end]
                        feed_dict = {self.x        : batch_x, 
                                     self.y        : batch_y, 
                                     self.keep_prob: keep_prob}
                        sess.run(minimizer, feed_dict=feed_dict)

                    # uncomment below to use custom accuracy calculations
                    # acc = self.accuracy(X_valid, y_valid, batch_size)
                    acc = round (self.measure([tf.metrics.accuracy], validation_data)[0], 3)
                    epoch, elapsed, hi_acc = self.afterEpoch(epoch, elapsed, 
                                                             acc, acc_save, hi_acc)
            
            except KeyboardInterrupt: # tip from https://stackoverflow.com/a/45033800
                self.logger.log("Manually terminated.")
        
        # tip from https://stackoverflow.com/a/2119512
        time_elapsed = datetime.now() - start_time
        hours = time_elapsed.seconds//3600
        mins  = time_elapsed.seconds//60 % 60
        secs  = time_elapsed.seconds % 60
        
        if hours > 0:
            msg = "...training complete. Highest accuracy: {:.3f}. Time elapsed: {}h,{}m,{}s."
            self.logger.log(msg.format(hi_acc, hours, mins, secs))        
        else:
            msg = "...training complete. Highest accuracy: {:.3f}. Time elapsed: {}m,{}s."
            self.logger.log(msg.format(hi_acc, mins, secs))        
        
        return
    
    def accHist(self):
        return self.acc_history
    
    # TODO: delete this?
    def test(self, test_data):
        
        X_test = test_data[0]
        y_test = test_data[1]
        
        with tf.Session() as sess:
            self.restore()
            score = self.measure([tf.metrics.accuracy], test_data, batch_size=128)
            msg = "Test accuracy: {:.3f}."
            self.logger.log(msg.format(score[0]))

            test_acc = self.accuracy((X_test, y_test), batch_size=129)
            msg = "Test accuracy non tf: {:.3f}."
            self.logger.log(msg.format(test_acc))

        return
    
    def metrics(self, data, metrics=[tf.metrics.accuracy,
                                     tf.metrics.precision,
                                     tf.metrics.recall]):
        with tf.Session():
            self.restore()
            scores = self.measure(metrics, data)
            
        metrics_msgs = ["Accuracy: {:.3f}", "Precision: {:.3f}", "Recall: {:.3f}"]
        for i, score in enumerate(scores):
            self.logger.log(metrics_msgs[i].format(score))
            
        return scores
    
    def metrics_by_class(self, data):

        with tf.Session():
            self.restore()

            self.logger.log("Calculating metrics by class...")
            metrics          = [tf.metrics.accuracy, tf.metrics.precision, tf.metrics.recall]
            accuracy_scores  = []
            precision_scores = []
            recall_scores    = []
            
            for classId in range(self.n_classes):
                scores = self.measure(metrics, data, classId)
                accuracy_scores.append(scores[0])
                precision_scores.append(scores[1])
                recall_scores.append(scores[2])

                msg = "ClassId {:>2} - Acc:{:.3f}, Pre:{:.3f}, Rcl:{:.3f}"
                self.logger.log(msg.format(classId, scores[0], scores[1], scores[2]))
                
        self.logger.log("...metrics calculated.")
            
        return (accuracy_scores, precision_scores, recall_scores)

    def predict(self, images, top=5):
        
        with tf.Session() as sess:
            
            self.restore()
            predictions = []
            
            self.logger.log("Classifying images...")
            
            op_top_k  = tf.nn.top_k(tf.nn.softmax(self.logits), top)
            feed_dict = {self.x        : images, 
                         self.keep_prob: 1.0}

            predictions = sess.run(op_top_k, feed_dict=feed_dict)

            msg = "...{} images classified."
            self.logger.log(msg.format(len(images)))
            
        return predictions
    
    def restore(self):

        saver = tf.train.Saver()
        sess  = tf.get_default_session()
        saver.restore(sess, "checkpoints/" + self.name)
        
        msg = "{} restored."
        self.logger.log(msg.format(self.name))        
        
        return
    
class Sequential(Model):
    
    def __init__(self, name, input_shape, n_classes):
        
        super().__init__(name, input_shape, n_classes)
        self.layers    = []
        
        return
        
    def addLayer(self, layer):
        
        self.layers.append(layer)
        
        return
    
    def assemble(self):
        
        prev_layer = self.input_layer
        for layer in self.layers:
            prev_layer = layer.connect(prev_layer)
            
        self.connectLogits(prev_layer)
        
        msg = "{} assembled."
        self.logger.log(msg.format(self.name))
            
        return 
    
    def summarize(self):
        
        msg = "     Summary for {}:"
        self.logger.log(msg.format(self.name))
        self.logger.log("----------------------------------")
        msg = "     Input".ljust(15) + ":{}"
        self.logger.log(msg.format(self.x.shape))
        self.logger.log("----------------------------------")
        
        n = 0
        for layer in self.layers:
            msg = "{:<2} : " + layer.summary()
            self.logger.log(msg.format(n))
            n += 1

        self.logger.log("----------------------------------")
        msg = "     Logits".ljust(15) + ":{}"
        self.logger.log(msg.format(self.logits.shape))
        self.logger.log("----------------------------------")
            
        return
    
    def eval_layer(self, i, img):
        
        msg = "Evaluating layer {}."
        self.logger.log(msg.format(i))
        
        tensor = self.layers[i].tensor
        feed_dict = {self.x        : [img], 
                     self.keep_prob: 1.0}

        with tf.Session() as sess:
            
            self.restore()
            eval_outputs = tensor.eval(session=sess, feed_dict=feed_dict)
            
        msg = "Output shape: {}."
        self.logger.log(msg.format(eval_outputs.shape))

        return eval_outputs

Model Architecture for Traffic Sign Classifier

The Sequential model is used to add the layers to develop the classifiers. Layers are added "sequentially" to the model.

The model architecture generally follows the LeNet architecture outlined in Lesson 13: Convolutional Neural Networks, Item 36. Lab: LeNet in Tensorflow.

A 1x1 convolution used in the very first layer has an output shape of 32x32x1. During development, this allowed the sequential model to improve validation accuracy as compared to training the model without it.

The other convolutional layers have the same first and second dimensions as the LeNet, but are deeper. The fully connected layers are wider with dropout added in the later stages.

tf.reset_default_graph()

notLenet = Sequential("notLeNet", input_shape=image_shape, n_classes=n_classes)

notLenet.addLayer (Conv2D   ([32, 32, 1]))
notLenet.addLayer (Conv2D   ([28, 28, 24]))
notLenet.addLayer (Pooling  ([14, 14]))
notLenet.addLayer (Conv2D   ([10, 10, 64]))
notLenet.addLayer (Pooling  ([5, 5]))
notLenet.addLayer (Flatten  ())
notLenet.addLayer (Connected([240]))
notLenet.addLayer (Dropout  ())    
notLenet.addLayer (Connected([168]))
notLenet.addLayer (Dropout  ())

notLenet.assemble()

W0211 22:25:20.439527  5528 deprecation.py:323] From C:\Users\Owner\anaconda3\envs\tsc\lib\site-packages\tensorflow_core\contrib\layers\python\layers\layers.py:1634: flatten (from tensorflow.python.layers.core) is deprecated and will be removed in a future version.
Instructions for updating:
Use keras.layers.flatten instead.
W0211 22:25:20.442496  5528 deprecation.py:323] From C:\Users\Owner\anaconda3\envs\tsc\lib\site-packages\tensorflow_core\python\layers\core.py:332: Layer.apply (from tensorflow.python.keras.engine.base_layer) is deprecated and will be removed in a future version.
Instructions for updating:
Please use `layer.__call__` method instead.
W0211 22:25:20.456461  5528 deprecation.py:506] From <ipython-input-17-4632de4b3f83>:71: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.
Instructions for updating:
Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.
W0211 22:25:20.484405  5528 deprecation.py:323] From <ipython-input-18-7fc2d30217b4>:37: softmax_cross_entropy_with_logits (from tensorflow.python.ops.nn_ops) is deprecated and will be removed in a future version.
Instructions for updating:

Future major versions of TensorFlow will allow gradients to flow
into the labels input on backprop by default.

See `tf.nn.softmax_cross_entropy_with_logits_v2`.

Model-2944 | notLeNet assembled.

Print Model Summary

Layers and respective output shapes are printed to verify architecture.

notLenet.summarize()

Model-2944 |      Summary for notLeNet:
Model-2944 | ----------------------------------
Model-2944 |      Input     :(?, 32, 32, 3)
Model-2944 | ----------------------------------
Model-2944 | 0  : Conv2D    :(?, 32, 32, 1)
Model-2944 | 1  : Conv2D    :(?, 28, 28, 24)
Model-2944 | 2  : Pooling   :(?, 14, 14, 24)
Model-2944 | 3  : Conv2D    :(?, 10, 10, 64)
Model-2944 | 4  : Pooling   :(?, 5, 5, 64)
Model-2944 | 5  : Flatten   :(?, 1600)
Model-2944 | 6  : Connected :(?, 240)
Model-2944 | 7  : Dropout   :(?, 240)
Model-2944 | 8  : Connected :(?, 168)
Model-2944 | 9  : Dropout   :(?, 168)
Model-2944 | ----------------------------------
Model-2944 |      Logits    :(?, 43)
Model-2944 | ----------------------------------

Train, Validate and Test the Model

A validation set can be used to assess how well the model is performing. A low accuracy on the training and validation sets imply underfitting. A high accuracy on the training set but low accuracy on the validation set implies overfitting.

### Train your model here.
### Calculate and report the accuracy on the training and validation set.
### Once a final model architecture is selected, 
### the accuracy on the test set should be calculated and reported as well.
### Feel free to use as many code cells as needed.

'''
Batch size of batch_size=128 was chosen through trial and error. 
Training terminates once epochs_done=64 epochs have passed 
since last highest validation accuracy was detected. 
If validation accuracy reaches acc_done=0.997, training will also terminate. 
Model is saved every time a new high accuracy is achieved. 
Keep probability for the dropout layers is keep_prob=0.5.
'''
notLenet.train(trainingSigns.data(), validSigns.data(), batch_size=128, 
        epochs_done=64, acc_done=0.997, keep_prob=0.5)

Model-2944 | Training...
Model-2944 | Timer started.
Model-2944 | Epoch    1 - Acc: 0.929
Model-2944 | Epoch    2 - New High Acc: 0.966
Model-2944 | notLeNet saved.
Model-2944 | Epoch    3 - New High Acc: 0.976
Model-2944 | notLeNet saved.
Model-2944 | Epoch    4 - New High Acc: 0.981
Model-2944 | notLeNet saved.
Model-2944 | Epoch    5 - Acc: 0.978 - Highest: 0.981
Model-2944 | Epoch    6 - Acc: 0.980 - Highest: 0.981
Model-2944 | Epoch    7 - Acc: 0.980 - Highest: 0.981
Model-2944 | Epoch    8 - Acc: 0.980 - Highest: 0.981
Model-2944 | Epoch    9 - New High Acc: 0.982
Model-2944 | notLeNet saved.
Model-2944 | Epoch   10 - New High Acc: 0.987
Model-2944 | notLeNet saved.
Model-2944 | Epoch   11 - Acc: 0.984 - Highest: 0.987
Model-2944 | Epoch   12 - Acc: 0.987 - Highest: 0.987
Model-2944 | Epoch   13 - Acc: 0.980 - Highest: 0.987
Model-2944 | Epoch   14 - Acc: 0.985 - Highest: 0.987
Model-2944 | Epoch   15 - Acc: 0.978 - Highest: 0.987
Model-2944 | Epoch   16 - Acc: 0.984 - Highest: 0.987
Model-2944 | Epoch   17 - Acc: 0.983 - Highest: 0.987
Model-2944 | Epoch   18 - Acc: 0.985 - Highest: 0.987
Model-2944 | Epoch   19 - Acc: 0.983 - Highest: 0.987
Model-2944 | Epoch   20 - Acc: 0.979 - Highest: 0.987
Model-2944 | Epoch   21 - Acc: 0.983 - Highest: 0.987
Model-2944 | Epoch   22 - Acc: 0.979 - Highest: 0.987
Model-2944 | Epoch   23 - Acc: 0.986 - Highest: 0.987
Model-2944 | Epoch   24 - Acc: 0.980 - Highest: 0.987
Model-2944 | Epoch   25 - Acc: 0.982 - Highest: 0.987
Model-2944 | Epoch   26 - Acc: 0.980 - Highest: 0.987
Model-2944 | Epoch   27 - Acc: 0.983 - Highest: 0.987
Model-2944 | Epoch   28 - Acc: 0.983 - Highest: 0.987
Model-2944 | Epoch   29 - Acc: 0.981 - Highest: 0.987
Model-2944 | Epoch   30 - Acc: 0.985 - Highest: 0.987
Model-2944 | Epoch   31 - Acc: 0.986 - Highest: 0.987
Model-2944 | Epoch   32 - New High Acc: 0.988
Model-2944 | notLeNet saved.
Model-2944 | Epoch   33 - Acc: 0.982 - Highest: 0.988
Model-2944 | Epoch   34 - Acc: 0.981 - Highest: 0.988
Model-2944 | Epoch   35 - Acc: 0.983 - Highest: 0.988
Model-2944 | Epoch   36 - Acc: 0.981 - Highest: 0.988
Model-2944 | Epoch   37 - Acc: 0.983 - Highest: 0.988
Model-2944 | Epoch   38 - Acc: 0.988 - Highest: 0.988
Model-2944 | Epoch   39 - Acc: 0.985 - Highest: 0.988
Model-2944 | Epoch   40 - Acc: 0.988 - Highest: 0.988
Model-2944 | Epoch   41 - Acc: 0.988 - Highest: 0.988
Model-2944 | Epoch   42 - Acc: 0.983 - Highest: 0.988
Model-2944 | Epoch   43 - Acc: 0.985 - Highest: 0.988
Model-2944 | Epoch   44 - Acc: 0.986 - Highest: 0.988
Model-2944 | Epoch   45 - Acc: 0.984 - Highest: 0.988
Model-2944 | Epoch   46 - Acc: 0.988 - Highest: 0.988
Model-2944 | Epoch   47 - Acc: 0.984 - Highest: 0.988
Model-2944 | Epoch   48 - Acc: 0.984 - Highest: 0.988
Model-2944 | Epoch   49 - New High Acc: 0.989
Model-2944 | notLeNet saved.
Model-2944 | Epoch   50 - Acc: 0.987 - Highest: 0.989
Model-2944 | Epoch   51 - Acc: 0.988 - Highest: 0.989
Model-2944 | Epoch   52 - Acc: 0.986 - Highest: 0.989
Model-2944 | Epoch   53 - Acc: 0.975 - Highest: 0.989
Model-2944 | Epoch   54 - Acc: 0.983 - Highest: 0.989
Model-2944 | Epoch   55 - Acc: 0.981 - Highest: 0.989
Model-2944 | Epoch   56 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch   57 - Acc: 0.978 - Highest: 0.989
Model-2944 | Epoch   58 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch   59 - Acc: 0.981 - Highest: 0.989
Model-2944 | Epoch   60 - Acc: 0.978 - Highest: 0.989
Model-2944 | Epoch   61 - Acc: 0.980 - Highest: 0.989
Model-2944 | Epoch   62 - Acc: 0.981 - Highest: 0.989
Model-2944 | Epoch   63 - Acc: 0.984 - Highest: 0.989
Model-2944 | Epoch   64 - Acc: 0.982 - Highest: 0.989
Model-2944 | Epoch   65 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch   66 - Acc: 0.980 - Highest: 0.989
Model-2944 | Epoch   67 - Acc: 0.984 - Highest: 0.989
Model-2944 | Epoch   68 - Acc: 0.981 - Highest: 0.989
Model-2944 | Epoch   69 - Acc: 0.984 - Highest: 0.989
Model-2944 | Epoch   70 - Acc: 0.987 - Highest: 0.989
Model-2944 | Epoch   71 - Acc: 0.982 - Highest: 0.989
Model-2944 | Epoch   72 - Acc: 0.983 - Highest: 0.989
Model-2944 | Epoch   73 - Acc: 0.986 - Highest: 0.989
Model-2944 | Epoch   74 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch   75 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch   76 - Acc: 0.982 - Highest: 0.989
Model-2944 | Epoch   77 - Acc: 0.982 - Highest: 0.989
Model-2944 | Epoch   78 - Acc: 0.986 - Highest: 0.989
Model-2944 | Epoch   79 - Acc: 0.984 - Highest: 0.989
Model-2944 | Epoch   80 - Acc: 0.980 - Highest: 0.989
Model-2944 | Epoch   81 - Acc: 0.979 - Highest: 0.989
Model-2944 | Epoch   82 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch   83 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch   84 - Acc: 0.986 - Highest: 0.989
Model-2944 | Epoch   85 - Acc: 0.986 - Highest: 0.989
Model-2944 | Epoch   86 - Acc: 0.986 - Highest: 0.989
Model-2944 | Epoch   87 - Acc: 0.986 - Highest: 0.989
Model-2944 | Epoch   88 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch   89 - Acc: 0.987 - Highest: 0.989
Model-2944 | Epoch   90 - Acc: 0.987 - Highest: 0.989
Model-2944 | Epoch   91 - Acc: 0.981 - Highest: 0.989
Model-2944 | Epoch   92 - Acc: 0.984 - Highest: 0.989
Model-2944 | Epoch   93 - Acc: 0.979 - Highest: 0.989
Model-2944 | Epoch   94 - Acc: 0.982 - Highest: 0.989
Model-2944 | Epoch   95 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch   96 - Acc: 0.986 - Highest: 0.989
Model-2944 | Epoch   97 - Acc: 0.987 - Highest: 0.989
Model-2944 | Epoch   98 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch   99 - Acc: 0.984 - Highest: 0.989
Model-2944 | Epoch  100 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch  101 - Acc: 0.983 - Highest: 0.989
Model-2944 | Epoch  102 - Acc: 0.978 - Highest: 0.989
Model-2944 | Epoch  103 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch  104 - Acc: 0.983 - Highest: 0.989
Model-2944 | Epoch  105 - Acc: 0.989 - Highest: 0.989
Model-2944 | Epoch  106 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch  107 - Acc: 0.985 - Highest: 0.989
Model-2944 | Epoch  108 - Acc: 0.984 - Highest: 0.989
Model-2944 | Epoch  109 - Acc: 0.987 - Highest: 0.989
Model-2944 | Epoch  110 - Acc: 0.983 - Highest: 0.989
Model-2944 | Epoch  111 - Acc: 0.981 - Highest: 0.989
Model-2944 | Epoch  112 - Acc: 0.984 - Highest: 0.989
Model-2944 | Epoch  113 - Acc: 0.984 - Highest: 0.989
Model-2944 | ...training complete. Highest accuracy: 0.989. Time elapsed: 20m,50s.

Plot Accuracy over Epochs

Model achieved a validation accuracy of 98.9% at epoch 49. Plot below shows performace over full range of epochs.

vwr.plotAccHist(notLenet.name, notLenet.accHist())

Plot Accuracy by Dataset

Model achieves 96.5% test accuracy.

train_acc = notLenet.metrics(trainingSigns.data(),metrics=[tf.metrics.accuracy])[0]
val_acc   = notLenet.metrics(validSigns.data(),metrics=[tf.metrics.accuracy])[0]
test_acc  = notLenet.metrics(testSigns.data(),metrics=[tf.metrics.accuracy])[0]

vwr.barhScores("notLenet Accuracy by Dataset", 
               [train_acc, val_acc, test_acc], 
               ["Training", "Validation", "Test"])

Model-2944 | notLeNet restored.
Model-2944 | Accuracy: 1.000
Model-2944 | notLeNet restored.
Model-2944 | Accuracy: 0.989
Model-2944 | notLeNet restored.
Model-2944 | Accuracy: 0.965

Calculate Performace Metrics by Traffic Sign Type

Caculate model's accuracy, precision and recall for a more detailed plot.

accuracy, precision, recall = notLenet.metrics_by_class(testSigns.data())

Model-2944 | notLeNet restored.
Model-2944 | Calculating metrics by class...

W0211 22:47:15.760925  5528 deprecation.py:323] From C:\Users\Owner\anaconda3\envs\tsc\lib\site-packages\tensorflow_core\python\ops\metrics_impl.py:2026: div (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Deprecated in favor of operator or tf.math.divide.
W0211 22:47:15.761922  5528 deprecation.py:323] From C:\Users\Owner\anaconda3\envs\tsc\lib\site-packages\tensorflow_core\python\ops\metrics_impl.py:2026: where (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.where in 2.0, which has the same broadcast rule as np.where

Model-2944 | ClassId  0 - Acc:1.000, Pre:1.000, Rcl:1.000
Model-2944 | ClassId  1 - Acc:0.998, Pre:0.992, Rcl:0.976
Model-2944 | ClassId  2 - Acc:0.997, Pre:0.948, Rcl:0.996
Model-2944 | ClassId  3 - Acc:0.996, Pre:0.957, Rcl:0.933
Model-2944 | ClassId  4 - Acc:0.997, Pre:0.978, Rcl:0.962
Model-2944 | ClassId  5 - Acc:0.993, Pre:0.919, Rcl:0.933
Model-2944 | ClassId  6 - Acc:0.998, Pre:1.000, Rcl:0.800
Model-2944 | ClassId  7 - Acc:0.998, Pre:0.973, Rcl:0.978
Model-2944 | ClassId  8 - Acc:0.997, Pre:0.950, Rcl:0.973
Model-2944 | ClassId  9 - Acc:0.999, Pre:0.988, Rcl:0.990
Model-2944 | ClassId 10 - Acc:0.999, Pre:0.997, Rcl:0.979
Model-2944 | ClassId 11 - Acc:0.997, Pre:0.952, Rcl:0.948
Model-2944 | ClassId 12 - Acc:0.999, Pre:0.990, Rcl:0.997
Model-2944 | ClassId 13 - Acc:1.000, Pre:0.997, Rcl:0.997
Model-2944 | ClassId 14 - Acc:1.000, Pre:1.000, Rcl:0.996
Model-2944 | ClassId 15 - Acc:0.998, Pre:0.913, Rcl:1.000
Model-2944 | ClassId 16 - Acc:1.000, Pre:0.993, Rcl:1.000
Model-2944 | ClassId 17 - Acc:1.000, Pre:1.000, Rcl:1.000
Model-2944 | ClassId 18 - Acc:0.995, Pre:0.980, Rcl:0.859
Model-2944 | ClassId 19 - Acc:1.000, Pre:0.968, Rcl:1.000
Model-2944 | ClassId 20 - Acc:0.998, Pre:0.756, Rcl:1.000
Model-2944 | ClassId 21 - Acc:0.998, Pre:0.874, Rcl:0.922
Model-2944 | ClassId 22 - Acc:1.000, Pre:1.000, Rcl:0.950
Model-2944 | ClassId 23 - Acc:0.999, Pre:0.909, Rcl:0.993
Model-2944 | ClassId 24 - Acc:0.999, Pre:0.956, Rcl:0.967
Model-2944 | ClassId 25 - Acc:0.997, Pre:0.960, Rcl:0.954
Model-2944 | ClassId 26 - Acc:0.997, Pre:0.880, Rcl:0.939
Model-2944 | ClassId 27 - Acc:0.997, Pre:0.673, Rcl:0.617
Model-2944 | ClassId 28 - Acc:1.000, Pre:0.980, Rcl:1.000
Model-2944 | ClassId 29 - Acc:0.999, Pre:0.946, Rcl:0.978
Model-2944 | ClassId 30 - Acc:0.997, Pre:0.870, Rcl:0.847
Model-2944 | ClassId 31 - Acc:0.998, Pre:0.988, Rcl:0.941
Model-2944 | ClassId 32 - Acc:0.999, Pre:0.882, Rcl:1.000
Model-2944 | ClassId 33 - Acc:1.000, Pre:0.981, Rcl:0.990
Model-2944 | ClassId 34 - Acc:1.000, Pre:1.000, Rcl:1.000
Model-2944 | ClassId 35 - Acc:0.997, Pre:0.997, Rcl:0.905
Model-2944 | ClassId 36 - Acc:1.000, Pre:0.968, Rcl:1.000
Model-2944 | ClassId 37 - Acc:0.999, Pre:0.857, Rcl:1.000
Model-2944 | ClassId 38 - Acc:0.999, Pre:0.988, Rcl:0.990
Model-2944 | ClassId 39 - Acc:1.000, Pre:0.957, Rcl:0.989
Model-2944 | ClassId 40 - Acc:0.999, Pre:0.875, Rcl:0.933
Model-2944 | ClassId 41 - Acc:1.000, Pre:0.951, Rcl:0.967
Model-2944 | ClassId 42 - Acc:1.000, Pre:0.989, Rcl:1.000
Model-2944 | ...metrics calculated.

Plot Accuracy by Training Sign

vwr.barhClassScores("notLenet Accuracy with Test Data by Sign (Low to High)", accuracy)

Plot Precision by Training Sign

vwr.barhClassScores("notLenet Precision with Test Data by Sign (Low to High)", precision)

Plot Recall by Training Sign

vwr.barhClassScores("notLenet Recall with Test Data by Sign (Low to High)", recall)

Most Difficult Signs

Rank the 10 hardest signs for model to recognize based on average score of accuracy, precision, and recall for type of sign.

vwr.barhCombinedClassScores("notLenet's 10 Hardest Signs (Average Score, Lowest to Highest)", 
                            accuracy, precision, recall, -10)

Most Easy Signs

Rank the 10 easiest signs for model to recognize based on average score of accuracy, precision, and recall for type of sign.

vwr.barhCombinedClassScores("notLenet's 10 Easiest Signs (Average Score, Highest to Lowest)", 
                            accuracy, precision, recall, 10)

---

Step 3: Test a Model on New Images

To give yourself more insight into how your model is working, download at least five pictures of German traffic signs from the web and use your model to predict the traffic sign type.

You may find signnames.csv useful as it contains mappings from the class id (integer) to the actual sign name.

Load and Output the Images

### Load the images and plot them here.
### Feel free to use as many code cells as needed.

# traffic signs were downloaded from wikipedia
folder_name = 'traffic_signs/wikipedia/'
wiki_traffic_signs = TrafficSignsDataset (folder_name, n_classes)
vwr.imShowImages(wiki_traffic_signs.X)

TrafficSignsDataset-6952 | Loading signs from traffic_signs/wikipedia/...
TrafficSignsDataset-6952 | Loaded 10,No passing for vehicles over 3.5 metric tons.png with label 10.
TrafficSignsDataset-6952 | Loaded 11,Right-of-way at the next intersection.png with label 11.
TrafficSignsDataset-6952 | Loaded 12,Priority road.png with label 12.
TrafficSignsDataset-6952 | Loaded 18,General caution.png with label 18.
TrafficSignsDataset-6952 | Loaded 19,Dangerous curve to the left.png with label 19.
TrafficSignsDataset-6952 | Loaded 20,Dangerous curve to the right.png with label 20.
TrafficSignsDataset-6952 | Loaded 21,Double curve.png with label 21.
TrafficSignsDataset-6952 | Loaded 22,Bumpy road.png with label 22.
TrafficSignsDataset-6952 | Loaded 24,Road narrows on the right.png with label 24.
TrafficSignsDataset-6952 | Loaded 26,Traffic signals.png with label 26.
TrafficSignsDataset-6952 | Loaded 27,Pedestrians.png with label 27.
TrafficSignsDataset-6952 | Loaded 29,Bicycles crossing.png with label 29.
TrafficSignsDataset-6952 | Loaded 3,Speed limit (60kmh).png with label 3.
TrafficSignsDataset-6952 | Loaded 30,Beware of ice-snow.png with label 30.
TrafficSignsDataset-6952 | Loaded 31,Wild animals crossing.png with label 31.
TrafficSignsDataset-6952 | Loaded 32,End of all speed and passing limits.png with label 32.
TrafficSignsDataset-6952 | Loaded 35,Ahead only.png with label 35.
TrafficSignsDataset-6952 | Loaded 40,Roundabout mandatory.png with label 40.
TrafficSignsDataset-6952 | Loaded 41,End of no passing.png with label 41.
TrafficSignsDataset-6952 | ...19 signs loaded.

C:\Users\Owner\anaconda3\envs\tsc\lib\site-packages\PIL\Image.py:963: UserWarning: Palette images with Transparency expressed in bytes should be converted to RGBA images
  "Palette images with Transparency expressed in bytes should be "

Predict the Sign Type for Each Image

### Run the predictions here and use the model to output the prediction for each image.
### Make sure to pre-process the images with the same pre-processing pipeline used earlier.
### Feel free to use as many code cells as needed.

wiki_traffic_signs.restoreOrgImages()
wiki_traffic_signs.processImages(image_processors)

TrafficSignsDataset-6952 | Original images and labels restored.
TrafficSignsDataset-6952 | Processing images...
TrafficSignsDataset-6952 | Applying equalizeHist() to images.
TrafficSignsDataset-6952 | Applying RGB_to_norm() to images.
TrafficSignsDataset-6952 | ...images processed.

Analyze Performance

### Calculate the accuracy for these 5 new images. 
### For example, if the model predicted 1 out of 5 signs correctly, it's 20% accurate on these new images.

notLenet_acc = notLenet.metrics(wiki_traffic_signs.data(), metrics=[tf.metrics.accuracy])[0]
vwr.barhScores("notLenet Accuracy with Traffic Signs from Wikipedia", [notLenet_acc], ["Accuracy"])

Model-2944 | notLeNet restored.
Model-2944 | Accuracy: 1.000

Output Top 5 Softmax Probabilities For Each Image Found on the Web

For each of the new images, print out the model's softmax probabilities to show the certainty of the model's predictions (limit the output to the top 5 probabilities for each image). tf.nn.top_k could prove helpful here.

The example below demonstrates how tf.nn.top_k can be used to find the top k predictions for each image.

tf.nn.top_k will return the values and indices (class ids) of the top k predictions. So if k=3, for each sign, it'll return the 3 largest probabilities (out of a possible 43) and the correspoding class ids.

Take this numpy array as an example. The values in the array represent predictions. The array contains softmax probabilities for five candidate images with six possible classes. tf.nn.top_k is used to choose the three classes with the highest probability:

# (5, 6) array
a = np.array([[ 0.24879643,  0.07032244,  0.12641572,  0.34763842,  0.07893497,
         0.12789202],
       [ 0.28086119,  0.27569815,  0.08594638,  0.0178669 ,  0.18063401,
         0.15899337],
       [ 0.26076848,  0.23664738,  0.08020603,  0.07001922,  0.1134371 ,
         0.23892179],
       [ 0.11943333,  0.29198961,  0.02605103,  0.26234032,  0.1351348 ,
         0.16505091],
       [ 0.09561176,  0.34396535,  0.0643941 ,  0.16240774,  0.24206137,
         0.09155967]])

Running it through sess.run(tf.nn.top_k(tf.constant(a), k=3)) produces:

TopKV2(values=array([[ 0.34763842,  0.24879643,  0.12789202],
       [ 0.28086119,  0.27569815,  0.18063401],
       [ 0.26076848,  0.23892179,  0.23664738],
       [ 0.29198961,  0.26234032,  0.16505091],
       [ 0.34396535,  0.24206137,  0.16240774]]), indices=array([[3, 0, 5],
       [0, 1, 4],
       [0, 5, 1],
       [1, 3, 5],
       [1, 4, 3]], dtype=int32))

Looking just at the first row we get [ 0.34763842, 0.24879643, 0.12789202], you can confirm these are the 3 largest probabilities in a. You'll also notice [3, 0, 5] are the corresponding indices.

### Print out the top five softmax probabilities for the predictions on the German traffic sign images found on the web. 
### Feel free to use as many code cells as needed.

predictions = notLenet.predict(wiki_traffic_signs.X, top=5)
vwr.imShowPredictions(notLenet.name, wiki_traffic_signs.org_X, predictions)

Model-2944 | notLeNet restored.
Model-2944 | Classifying images...
Model-2944 | ...19 images classified.

Project Writeup

Once you have completed the code implementation, document your results in a project writeup using this template as a guide. The writeup can be in a markdown or pdf file.

Note: Once you have completed all of the code implementations and successfully answered each question above, you may finalize your work by exporting the iPython Notebook as an HTML document. You can do this by using the menu above and navigating to \n", "File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.

---

Step 4 (Optional): Visualize the Neural Network's State with Test Images

This Section is not required to complete but acts as an additional excersise for understaning the output of a neural network's weights. While neural networks can be a great learning device they are often referred to as a black box. We can understand what the weights of a neural network look like better by plotting their feature maps. After successfully training your neural network you can see what it's feature maps look like by plotting the output of the network's weight layers in response to a test stimuli image. From these plotted feature maps, it's possible to see what characteristics of an image the network finds interesting. For a sign, maybe the inner network feature maps react with high activation to the sign's boundary outline or to the contrast in the sign's painted symbol.

Provided for you below is the function code that allows you to get the visualization output of any tensorflow weight layer you want. The inputs to the function should be a stimuli image, one used during training or a new one you provided, and then the tensorflow variable name that represents the layer's state during the training process, for instance if you wanted to see what the LeNet lab's feature maps looked like for it's second convolutional layer you could enter conv2 as the tf_activation variable.

For an example of what feature map outputs look like, check out NVIDIA's results in their paper End-to-End Deep Learning for Self-Driving Cars in the section Visualization of internal CNN State. NVIDIA was able to show that their network's inner weights had high activations to road boundary lines by comparing feature maps from an image with a clear path to one without. Try experimenting with a similar test to show that your trained network's weights are looking for interesting features, whether it's looking at differences in feature maps from images with or without a sign, or even what feature maps look like in a trained network vs a completely untrained one on the same sign image.

Your output should look something like this (above)

### Visualize your network's feature maps here.
### Feel free to use as many code cells as needed.

# image_input: the test image being fed into the network to produce the feature maps
# tf_activation: should be a tf variable name used during your training procedure that represents the calculated state of a specific weight layer
# activation_min/max: can be used to view the activation contrast in more detail, by default matplot sets min and max to the actual min and max values of the output
# plt_num: used to plot out multiple different weight feature map sets on the same block, just extend the plt number for each new feature map entry

def outputFeatureMap(image_input, tf_activation, activation_min=-1, activation_max=-1 ,plt_num=1):
    # Here make sure to preprocess your image_input in a way your network expects
    # with size, normalization, ect if needed
    # image_input =
    # Note: x should be the same name as your network's tensorflow data placeholder variable
    # If you get an error tf_activation is not defined it may be having trouble accessing the variable from inside a function
    activation = tf_activation.eval(session=sess,feed_dict={x : image_input})
    featuremaps = activation.shape[3]
    plt.figure(plt_num, figsize=(15,15))
    for featuremap in range(featuremaps):
        plt.subplot(6,8, featuremap+1) # sets the number of feature maps to show on each row and column
        plt.title('FeatureMap ' + str(featuremap)) # displays the feature map number
        if activation_min != -1 & activation_max != -1:
            plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", vmin =activation_min, vmax=activation_max, cmap="gray")
        elif activation_max != -1:
            plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", vmax=activation_max, cmap="gray")
        elif activation_min !=-1:
            plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", vmin=activation_min, cmap="gray")
        else:
            plt.imshow(activation[0,:,:, featuremap], interpolation="nearest", cmap="gray")

Select and Preprocess Image for Network State Visualization

x = trainingSigns.org_X[8500].copy()

equ = equalizeHist(x)
img = RGB_to_norm(equ)

plt.imshow(norm_to_RGB(img))

<matplotlib.image.AxesImage at 0x245eaf91b00>

Generate Layer 0 Feature Map with Image

layer_outputs = notLenet.eval_layer(0, img)
vwr.imShowConvOutputs(layer_outputs)

Model-2944 | Evaluating layer 0.
Model-2944 | notLeNet restored.
Model-2944 | Output shape: (1, 32, 32, 1).

Generate Layer 1 Feature Map with Image

layer_outputs = notLenet.eval_layer(1, img)
vwr.imShowConvOutputs(layer_outputs)

Model-2944 | Evaluating layer 1.
Model-2944 | notLeNet restored.
Model-2944 | Output shape: (1, 28, 28, 24).

---

Out of Curiosity: Attempt to Replicate Multi-Scale CNN

Just out of curiosity, a multi-scale CNN is attempted.

Concatenation Layer

A layer that concatenates flattened layers is required.

class Concatenate(Layer):
    
    def setName(self):
        self.name = "Concatenate"
        return

    def connect(self, *prev_layers):
        self.model = prev_layers[0].model
        tensors = [layer.tensor for layer in prev_layers]
        self.tensor = tf.concat(tensors, axis=1)
        return self

Create Multi-Scale Architecture

The architecture is similar to the one Sermanet and Lecunn described in "Traffic Sign Recognition with Multi-Scale Convolutional Networks" where the output of the first stage is fed to the classifier stage.

notSermanet = Model("notSermanet", image_shape, n_classes)

stage_1x1 = Conv2D ([32, 32, 1]).connect(notSermanet.inputLayer())

stage_1 = Conv2D ([28, 28, 24]).connect(stage_1x1)
stage_1 = Pooling([14, 14]).connect(stage_1)
flatten_1 = Flatten().connect(stage_1)

stage_2 = Conv2D ([10, 10, 64]).connect(stage_1)
stage_2 = Pooling([ 5,  5]).connect(stage_2)
stage_2 = Conv2D ([3, 3, 96]).connect(stage_2)
flatten_2 = Flatten().connect(stage_2)

stage_3 = Concatenate().connect(flatten_1, flatten_2)
stage_3 = Connected([240]).connect(stage_3)
stage_3 = Dropout().connect(stage_3)
stage_3 = Connected([168]).connect(stage_3)
stage_3 = Dropout().connect(stage_3)

notSermanet.connectLogits(stage_3)

Train Multi-Scale CNN

notSermanet.train(trainingSigns.data(), validSigns.data(), batch_size=128, 
        epochs_done=64, acc_done=0.997, keep_prob=0.5)

Model-6552 | Training...
Model-6552 | Timer started.
Model-6552 | Epoch    1 - New High Acc: 0.964
Model-6552 | notSermanet saved.
Model-6552 | Epoch    2 - New High Acc: 0.979
Model-6552 | notSermanet saved.
Model-6552 | Epoch    3 - New High Acc: 0.983
Model-6552 | notSermanet saved.
Model-6552 | Epoch    4 - Acc: 0.980 - Highest: 0.983
Model-6552 | Epoch    5 - Acc: 0.981 - Highest: 0.983
Model-6552 | Epoch    6 - Acc: 0.982 - Highest: 0.983
Model-6552 | Epoch    7 - Acc: 0.981 - Highest: 0.983
Model-6552 | Epoch    8 - Acc: 0.980 - Highest: 0.983
Model-6552 | Epoch    9 - New High Acc: 0.984
Model-6552 | notSermanet saved.
Model-6552 | Epoch   10 - Acc: 0.980 - Highest: 0.984
Model-6552 | Epoch   11 - Acc: 0.980 - Highest: 0.984
Model-6552 | Epoch   12 - New High Acc: 0.987
Model-6552 | notSermanet saved.
Model-6552 | Epoch   13 - Acc: 0.978 - Highest: 0.987
Model-6552 | Epoch   14 - Acc: 0.987 - Highest: 0.987
Model-6552 | Epoch   15 - Acc: 0.984 - Highest: 0.987
Model-6552 | Epoch   16 - Acc: 0.982 - Highest: 0.987
Model-6552 | Epoch   17 - Acc: 0.981 - Highest: 0.987
Model-6552 | Epoch   18 - Acc: 0.985 - Highest: 0.987
Model-6552 | Epoch   19 - Acc: 0.985 - Highest: 0.987
Model-6552 | Epoch   20 - Acc: 0.979 - Highest: 0.987
Model-6552 | Epoch   21 - Acc: 0.980 - Highest: 0.987
Model-6552 | Epoch   22 - Acc: 0.987 - Highest: 0.987
Model-6552 | Epoch   23 - Acc: 0.984 - Highest: 0.987
Model-6552 | Epoch   24 - Acc: 0.987 - Highest: 0.987
Model-6552 | Epoch   25 - Acc: 0.982 - Highest: 0.987
Model-6552 | Epoch   26 - Acc: 0.987 - Highest: 0.987
Model-6552 | Epoch   27 - Acc: 0.982 - Highest: 0.987
Model-6552 | Epoch   28 - Acc: 0.985 - Highest: 0.987
Model-6552 | Epoch   29 - Acc: 0.984 - Highest: 0.987
Model-6552 | Epoch   30 - Acc: 0.982 - Highest: 0.987
Model-6552 | Epoch   31 - Acc: 0.984 - Highest: 0.987
Model-6552 | Epoch   32 - Acc: 0.973 - Highest: 0.987
Model-6552 | Epoch   33 - Acc: 0.979 - Highest: 0.987
Model-6552 | Epoch   34 - Acc: 0.983 - Highest: 0.987
Model-6552 | Epoch   35 - Acc: 0.978 - Highest: 0.987
Model-6552 | Epoch   36 - Acc: 0.986 - Highest: 0.987
Model-6552 | Epoch   37 - Acc: 0.987 - Highest: 0.987
Model-6552 | Epoch   38 - Acc: 0.981 - Highest: 0.987
Model-6552 | Epoch   39 - Acc: 0.985 - Highest: 0.987
Model-6552 | Epoch   40 - Acc: 0.978 - Highest: 0.987
Model-6552 | Epoch   41 - Acc: 0.978 - Highest: 0.987
Model-6552 | Epoch   42 - New High Acc: 0.990
Model-6552 | notSermanet saved.
Model-6552 | Epoch   43 - Acc: 0.985 - Highest: 0.990
Model-6552 | Epoch   44 - Acc: 0.980 - Highest: 0.990
Model-6552 | Epoch   45 - Acc: 0.982 - Highest: 0.990
Model-6552 | Epoch   46 - Acc: 0.985 - Highest: 0.990
Model-6552 | Epoch   47 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch   48 - Acc: 0.984 - Highest: 0.990
Model-6552 | Epoch   49 - Acc: 0.977 - Highest: 0.990
Model-6552 | Epoch   50 - Acc: 0.985 - Highest: 0.990
Model-6552 | Epoch   51 - Acc: 0.985 - Highest: 0.990
Model-6552 | Epoch   52 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch   53 - Acc: 0.980 - Highest: 0.990
Model-6552 | Epoch   54 - Acc: 0.982 - Highest: 0.990
Model-6552 | Epoch   55 - Acc: 0.976 - Highest: 0.990
Model-6552 | Epoch   56 - Acc: 0.977 - Highest: 0.990
Model-6552 | Epoch   57 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch   58 - Acc: 0.980 - Highest: 0.990
Model-6552 | Epoch   59 - Acc: 0.977 - Highest: 0.990
Model-6552 | Epoch   60 - Acc: 0.975 - Highest: 0.990
Model-6552 | Epoch   61 - Acc: 0.976 - Highest: 0.990
Model-6552 | Epoch   62 - Acc: 0.982 - Highest: 0.990
Model-6552 | Epoch   63 - Acc: 0.981 - Highest: 0.990
Model-6552 | Epoch   64 - Acc: 0.979 - Highest: 0.990
Model-6552 | Epoch   65 - Acc: 0.982 - Highest: 0.990
Model-6552 | Epoch   66 - Acc: 0.985 - Highest: 0.990
Model-6552 | Epoch   67 - Acc: 0.979 - Highest: 0.990
Model-6552 | Epoch   68 - Acc: 0.978 - Highest: 0.990
Model-6552 | Epoch   69 - Acc: 0.978 - Highest: 0.990
Model-6552 | Epoch   70 - Acc: 0.979 - Highest: 0.990
Model-6552 | Epoch   71 - Acc: 0.980 - Highest: 0.990
Model-6552 | Epoch   72 - Acc: 0.984 - Highest: 0.990
Model-6552 | Epoch   73 - Acc: 0.979 - Highest: 0.990
Model-6552 | Epoch   74 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch   75 - Acc: 0.988 - Highest: 0.990
Model-6552 | Epoch   76 - Acc: 0.985 - Highest: 0.990
Model-6552 | Epoch   77 - Acc: 0.988 - Highest: 0.990
Model-6552 | Epoch   78 - Acc: 0.980 - Highest: 0.990
Model-6552 | Epoch   79 - Acc: 0.981 - Highest: 0.990
Model-6552 | Epoch   80 - Acc: 0.978 - Highest: 0.990
Model-6552 | Epoch   81 - Acc: 0.979 - Highest: 0.990
Model-6552 | Epoch   82 - Acc: 0.986 - Highest: 0.990
Model-6552 | Epoch   83 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch   84 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch   85 - Acc: 0.977 - Highest: 0.990
Model-6552 | Epoch   86 - Acc: 0.984 - Highest: 0.990
Model-6552 | Epoch   87 - Acc: 0.988 - Highest: 0.990
Model-6552 | Epoch   88 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch   89 - Acc: 0.979 - Highest: 0.990
Model-6552 | Epoch   90 - Acc: 0.970 - Highest: 0.990
Model-6552 | Epoch   91 - Acc: 0.977 - Highest: 0.990
Model-6552 | Epoch   92 - Acc: 0.980 - Highest: 0.990
Model-6552 | Epoch   93 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch   94 - Acc: 0.986 - Highest: 0.990
Model-6552 | Epoch   95 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch   96 - Acc: 0.986 - Highest: 0.990
Model-6552 | Epoch   97 - Acc: 0.971 - Highest: 0.990
Model-6552 | Epoch   98 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch   99 - Acc: 0.979 - Highest: 0.990
Model-6552 | Epoch  100 - Acc: 0.978 - Highest: 0.990
Model-6552 | Epoch  101 - Acc: 0.982 - Highest: 0.990
Model-6552 | Epoch  102 - Acc: 0.982 - Highest: 0.990
Model-6552 | Epoch  103 - Acc: 0.973 - Highest: 0.990
Model-6552 | Epoch  104 - Acc: 0.983 - Highest: 0.990
Model-6552 | Epoch  105 - Acc: 0.981 - Highest: 0.990
Model-6552 | Epoch  106 - Acc: 0.983 - Highest: 0.990
Model-6552 | ...training complete. Highest accuracy: 0.990. Time elapsed: 24m,57s.

Plot Accuracy over Epochs

vwr.plotAccHist(notSermanet.name, notSermanet.accHist())

Plot Accuracy by Dataset

train_acc = notSermanet.metrics(trainingSigns.data(),metrics=[tf.metrics.accuracy])[0]
val_acc   = notSermanet.metrics(validSigns.data(),metrics=[tf.metrics.accuracy])[0]
test_acc  = notSermanet.metrics(testSigns.data(),metrics=[tf.metrics.accuracy])[0]

vwr.barhScores("notSermanet Accuracy by Dataset", 
               [train_acc, val_acc, test_acc], 
               ["Training", "Validation", "Test"])

Model-6552 | notSermanet restored.
Model-6552 | Accuracy: 1.000
Model-6552 | notSermanet restored.
Model-6552 | Accuracy: 0.990
Model-6552 | notSermanet restored.
Model-6552 | Accuracy: 0.971

Calculate Performace Metrics by Traffic Sign Type

notSermanet_acc, notSermanet_pre, notSermanet_rcl = notSermanet.metrics_by_class(testSigns.data())

Model-6552 | notSermanet restored.
Model-6552 | Calculating metrics by class...
Model-6552 | ClassId  0 - Acc:1.000, Pre:1.000, Rcl:0.967
Model-6552 | ClassId  1 - Acc:0.999, Pre:0.981, Rcl:0.994
Model-6552 | ClassId  2 - Acc:0.998, Pre:0.984, Rcl:0.991
Model-6552 | ClassId  3 - Acc:0.997, Pre:0.952, Rcl:0.962
Model-6552 | ClassId  4 - Acc:0.998, Pre:0.988, Rcl:0.979
Model-6552 | ClassId  5 - Acc:0.995, Pre:0.923, Rcl:0.983
Model-6552 | ClassId  6 - Acc:0.997, Pre:1.000, Rcl:0.780
Model-6552 | ClassId  7 - Acc:0.998, Pre:0.969, Rcl:0.984
Model-6552 | ClassId  8 - Acc:0.998, Pre:0.973, Rcl:0.969
Model-6552 | ClassId  9 - Acc:0.998, Pre:0.961, Rcl:0.988
Model-6552 | ClassId 10 - Acc:0.999, Pre:1.000, Rcl:0.986
Model-6552 | ClassId 11 - Acc:0.995, Pre:0.979, Rcl:0.869
Model-6552 | ClassId 12 - Acc:0.999, Pre:0.990, Rcl:0.991
Model-6552 | ClassId 13 - Acc:1.000, Pre:1.000, Rcl:0.997
Model-6552 | ClassId 14 - Acc:1.000, Pre:1.000, Rcl:1.000
Model-6552 | ClassId 15 - Acc:0.999, Pre:0.967, Rcl:0.990
Model-6552 | ClassId 16 - Acc:1.000, Pre:1.000, Rcl:1.000
Model-6552 | ClassId 17 - Acc:1.000, Pre:0.997, Rcl:1.000
Model-6552 | ClassId 18 - Acc:0.995, Pre:1.000, Rcl:0.846
Model-6552 | ClassId 19 - Acc:1.000, Pre:0.984, Rcl:1.000
Model-6552 | ClassId 20 - Acc:0.998, Pre:0.795, Rcl:0.989
Model-6552 | ClassId 21 - Acc:1.000, Pre:0.977, Rcl:0.956
Model-6552 | ClassId 22 - Acc:1.000, Pre:1.000, Rcl:0.983
Model-6552 | ClassId 23 - Acc:0.997, Pre:0.819, Rcl:0.967
Model-6552 | ClassId 24 - Acc:0.999, Pre:0.988, Rcl:0.922
Model-6552 | ClassId 25 - Acc:0.999, Pre:0.998, Rcl:0.969
Model-6552 | ClassId 26 - Acc:0.998, Pre:0.898, Rcl:0.983
Model-6552 | ClassId 27 - Acc:0.998, Pre:0.738, Rcl:0.983
Model-6552 | ClassId 28 - Acc:0.999, Pre:0.986, Rcl:0.947
Model-6552 | ClassId 29 - Acc:0.999, Pre:0.856, Rcl:0.989
Model-6552 | ClassId 30 - Acc:0.996, Pre:0.798, Rcl:0.893
Model-6552 | ClassId 31 - Acc:0.998, Pre:0.930, Rcl:0.978
Model-6552 | ClassId 32 - Acc:1.000, Pre:0.968, Rcl:1.000
Model-6552 | ClassId 33 - Acc:1.000, Pre:0.977, Rcl:0.995
Model-6552 | ClassId 34 - Acc:1.000, Pre:0.952, Rcl:1.000
Model-6552 | ClassId 35 - Acc:0.999, Pre:0.990, Rcl:0.967
Model-6552 | ClassId 36 - Acc:1.000, Pre:0.992, Rcl:1.000
Model-6552 | ClassId 37 - Acc:1.000, Pre:1.000, Rcl:0.983
Model-6552 | ClassId 38 - Acc:0.999, Pre:0.999, Rcl:0.983
Model-6552 | ClassId 39 - Acc:1.000, Pre:0.989, Rcl:0.956
Model-6552 | ClassId 40 - Acc:1.000, Pre:1.000, Rcl:0.944
Model-6552 | ClassId 41 - Acc:0.999, Pre:1.000, Rcl:0.817
Model-6552 | ClassId 42 - Acc:1.000, Pre:1.000, Rcl:0.967
Model-6552 | ...metrics calculated.

Accuracy by Traffic Sign

vwr.barhClassScores("notSermanet Accuracy with Test Data by Sign", notSermanet_acc)

Precision by Traffic Sign

vwr.barhClassScores("notSermanet Precision with Test Data by Sign", notSermanet_pre)

Recall by Traffic Sign

vwr.barhClassScores("notSermanet Recall with Test Data by Sign", notSermanet_rcl)

Most Difficult Signs

Rank the 10 hardest signs for model to recognize based on average score of accuracy, precision, and recall for type of sign.

vwr.barhCombinedClassScores("notSermanet's 10 Hardest Signs (Average Score, Lowest to Highest)", 
                            notSermanet_acc, notSermanet_pre, notSermanet_rcl, -10)

Most Easy Signs

Rank the 10 easiest signs for model to recognize based on average score of accuracy, precision, and recall for type of sign.

vwr.barhCombinedClassScores("notSermanet's 10 Easiest Signs (Average Score, Highest to Lowest)", 
                            notSermanet_acc, notSermanet_pre, notSermanet_rcl, 10)

Check Performance with Downloaded Traffic Signs

accuracy = notSermanet.metrics(wiki_traffic_signs.data(), metrics=[tf.metrics.accuracy])[0]
vwr.barhScores("notSermanet Accuracy with Wikipedia German Traffic Signs", [accuracy], ["Accuracy"])

Model-6552 | notSermanet restored.
Model-6552 | Accuracy: 1.000

View Model Classifications of Downloaded Traffic Signs

predictions = notSermanet.predict(wiki_traffic_signs.X, top=5)
vwr.imShowPredictions(notSermanet.name, wiki_traffic_signs.org_X, predictions)

Model-6552 | notSermanet restored.
Model-6552 | Classifying images...
Model-6552 | ...19 images classified.